from collections import defaultdict
from utils.landmarks import load_landmark_model, perform_cmr_landmark_detection
from utils.cfg import load_config
from utils.transforms import get_segmentation_transforms
from utils.inference import center_crop, pad_if_needed
from utils.vis import compute_bullseye_sector_mask_for_slice
LANDMARK_MODELPATH = "E:/Data/T1T2_models/CMR_landmark_network_RO_352_E1_352_sax_with_T1_T1T2_LGE_PerfPD_LossMultiSoftProb_KLD_Dice_Pytorch_1.5.1_2020-08-13_20200813_181146.pts"
LABEL_DIRS = [r"E:\Data\T1T2_peter_test_james", r"E:\Data\T1T2_peter_test_hui"]
CONFIG = "./experiments/026.yaml"
FOV = 256
WRITE_PNGS = True
# Load config
cfg, model_dir = load_config(CONFIG)
sequences = cfg['export']['sequences']
label_classes = cfg['export']['label_classes']
gaussian_sigma = cfg['export']['gaussian_sigma']
n_channels_keep_img = len(cfg['export']['sequences']) # May have exported more channels to make PNG
n_channels_keep_lab = len(cfg['export']['label_classes'])
# landmark model
landmark_model = load_landmark_model(LANDMARK_MODELPATH)
_, transforms_test = get_segmentation_transforms(cfg)
for label_dir in LABEL_DIRS:
out_dir = os.path.join("./data/", os.path.basename(label_dir))
if not os.path.exists(out_dir):
if __name__ == '__main__':
# parse arguments
parser = argparse.ArgumentParser()
parser.add_argument('--data_path',
type=str,
default='data/test_03_temp/multiview_data')
parser.add_argument('--output_folder',
type=str,
default='results/test_03_temp')
args = parser.parse_args()
# prepare model and data
device = torch.device(0)
config = cfg.load_config('experiments/syn_data/multiview_data_2_alg.yaml')
model = AlgebraicTriangulationNet(config, device=device).to(device)
state_dict = torch.load(config.model.checkpoint)
for key in list(state_dict.keys()):
new_key = key.replace("module.", "")
state_dict[new_key] = state_dict.pop(key)
model.load_state_dict(state_dict, strict=True)
print("Loading data ...")
dataset = MultiView_SynData(args.data_path,
invalid_joints=(9, 16),
bbox=[80, 0, 560, 480],
ori_form=1)
dataloader = datasets_utils.syndata_loader(dataset, batch_size=4)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
'--config',
type=str,
default="experiments/human36m/train/human36m_alg_17jnts.yaml")
parser.add_argument('--resume', action='store_true')
parser.add_argument('--logdir', type=str, default="./logs")
parser.add_argument('--gamma', type=float, default=10.0)
parser.add_argument('--resume_log', type=str, default="")
args = parser.parse_args()
config = cfg.load_config(args.config)
assert config.dataset.type in ("syndata", "human36m", "mpii")
device = torch.device(int(config.gpu_id))
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)
model = AlgebraicTriangulationNet(config, device=device)
model = torch.nn.DataParallel(model, device_ids=[int(config.gpu_id)])
if config.model.init_weights:
print("Initializing model weights..")
if args.resume:
last_epoch = int(
import os
import multiprocessing
from glob import glob
from tqdm import tqdm
from utils.cfg import load_config
from utils.export import export_label
CONFIG = "../experiments/026.yaml"
EXCLUDED_FILES_PATH = "data/blacklist.txt"
# Load config
cfg, vis_dir, model_dir = load_config(CONFIG)
npy_dir = cfg['export']['npydir']
output_dir = os.path.join(cfg['data']['pngdir'])
sequences = cfg['export']['sequences']
label_classes = cfg['export']['label_classes']
gaussian_sigma = cfg['export']['gaussian_sigma']
frmt = cfg['export']['format']
# Excluded files
with open(EXCLUDED_FILES_PATH) as f:
excluded_files = f.read().splitlines()
def export_label_helper(labelpath):
export_label(labelpath, frmt, sequences, label_classes, output_dir,
gaussian_sigma)
